Bots are officiallyeverywhere — they’re available to assist with almost every part of our lives, from shopping and gift-giving to news-tracking and political decision-making. They can even tell us what to make for dinner .
The thing that seems odd to me, though, is that the entire bot conversation has been focused on A.I., machine learning (ML), and natural language processing (NLP). This makes me question whether we’re looking at bots the wrong way.
Having been in the business of building software for quite some time, and learned about workflows, business logic and other riveting concepts, I can’t help but feel there’s a misunderstanding about what constitutes a bot and what is just an access layer to the bot.
The buzzwords (AI, ML, NLP) are all additional interaction methods that sit further up the stack than the actual bot technology. While they can all be used to benefit the underlying bot and simplify access, it’s really the bot infrastructure itself that is the most complex and interesting. The parts missing from the macro conversation are:
What does the bot do? Why does it do it? What else can it do? How can it do that?
One example of a company using bots in an innovative way is WSC Sports Technologies . Their bot-based platform analyzes sports broadcasts in real time, identifies every component in a game, and generates customized highlights based on any element. The value proposition could require a bunch of deep A.I. or machine learning integration in order to work, but it’s actually based on allowing the tracking and rule-based outputs that are desirable for the NBA. It’s processing this as a data transaction, not A.I.
Another example that is particularly timely comes from a company called SapientX, which has createdchatbots for Hillary Clinton and Donald Trump that respond to policy questions users ask based on the candidate’s previous statements. Is that A.I.? It is taking the information and presenting it in a new way that’s useful.
I’d actually contend that one of the biggest opportunities for bots to make an impact today is by simplifying time-consuming processes in the workplace.
Take, for example, the process a shoestore manager goes through to reorder inventory. She’s having a conversation via walkie-talkie with her assistant manager, who mentions they are out of gold wedge sandals in size 7.
In order to check the inventory and reorder and schedule delivery, the manager would normally need to log into multiple backend systems to check information and perform actions — usually taking up a POS or getting off the floor to use a back room computer. But, with a bot for Slack that’s programmed to check the inventory and give her the option to reorder directly within her conversation, she saves a lot of time and effort, and the assistant manager is automatically updated on the status of the situation.
And imagine how much time and money would be saved if this became the norm across the entire organization. If thousands of employees across hundreds of stores are saving a few minutes on a process they probably complete several times a day, that would be extremely significant.
So, to recap, there are really two fairly distinct parts of the bot business. There’s the actual underlying bot programming and the thinking and work it takes to set up the syntax of each use case, and then there’s the additional layers that incorporate machine learning or NLP, which are really just ways to access and interact with the bots more easily.
The companies working on the interaction layer at the moment actually make a lot of sense. Google has its Cloud Platform for machine learning and A.I., Apple has Siri and its corresponding API, and Microsoft has Cortana. All of these companies have massive engines and pools of data and information to help their platforms get better and learn faster.
The part that’s extremely interesting to me is understanding what we’re going to use bots for in our everyday lives. Creating a bot just to view customer data in Salesforce is of pretty limited value in the long term; it’s similar to business dashboarding on mobile, where users can only pull information, not actually perform actions. The most useful bots will allow us to complete actions and entire processes more quickly and easily.
And the ecosystem will only continue to expand. Bots are at an inflection point similar to where apps were about 10 years ago. They’re attracting the attention of developers and consumers, and large companies are racing into the space in hopes of leading the charge and becoming the brokers of the bots, the same way Apple and Google did with apps.
The real opportunity right now for companies trying to get into bots is to think of the killer use cases and build them. How are bots going to become a business mainstay we can’t live without? That’s the question I think we should be discussing.